Missing json_encoders in v2 - Any similar functionality

[sachingiridhar-acelr]

This is just a small example of my use case. I have a RootModel which I use across my project. I have to json encode this at one place to avoid duplication of code (DRY).

from datetime import time as time_only
from pydantic import BaseModel, Field

def convert_time_to_hh_mm_ss(dt: time_only) -> str:
    return dt.strftime('%H:%M:%S')

class RootModel(BaseModel):
    class Config:
        json_encoders = {
            time_only: convert_time_to_hh_mm_ss
        }

I can see that json_encoders are removed in Pydantic v2. So how do I achieve the same in v2. This is very important to me since I have around 20 to 30 different models which needs this functionality

[xmatthias]

Facing a similar problem in #6726
while my problem is not directly applicable for a root-base model - it's still a quite similar problem.

[eyemyth]

I believe the new approach would be to use the functional serializers

[samuelcolvin]

Yes, functional serializers are the proper solution, but we've just added support (albeit with a deprecation warning) back in #6811

[voxsim]

@samuelcolvin probably makes sense to have somewhere an example that translates from the v1 to v2 because it is not super clear from the documentation.

[Samreay]

I'd be keen to get an official example as well.

Our use case is fairly simple, in that our pydantic models have some pandas dataframes as attritubtes, so we have json_encoders={pd.DataFrame=lambda x: x.reset_index().to_dict(orient="list")}.

Given this applies to all dataframe attritbutes without having to write out the field name for all of them, its very handy.

But with the new approach, I'm not sure if I should be using field_serializer, model_serialiser, PlainSerializer or WrapSerializer.

[alex-janss]

Here's an example using PlainSerializer that i used.
instead of:

class Foo(BaseModel):
    model_config = ConfigDict(json_encoders={str: lambda x: ' '.join(x)})
    x: str
    y: str = Field(...)

create a custom type:

CustomStr = Annotated[str, PlainSerializer(lambda x: ' '.join(x))]
class Foo(BaseModel):
    x: CustomStr
    y: CustomStr = Field(...)

[cherrywins]

How eventually have you rewritten your code?

[devamin]

I got this from the test written in the feature support PR :

    class Model(BaseModel):
        x: timedelta
        y: Decimal
        z: date

        model_config = ConfigDict(
            json_encoders={timedelta: lambda v: f'{v.total_seconds():0.3f}s', Decimal: lambda v: 'a decimal'}
        )